Helping people learn more from your data visualisations.
Visualising data is a key part of Vizzuality’s projects. Every day we try to figure out how we can make our visualisations better, so people that read them learn about the world. We also want to make sure people interpret data accurately and don’t make false conclusions. Especially when the future of the planet is on the line, small misinterpretations could make a big difference.
There’s lots of ways to help people make the most of your data visualisations. In this post I want to outline a number of things you can do to challenge people’s minds and ensure they build a great understanding of the world around us.
An example of misinterpretation in action.
In my last post, I talked about how people can misinterpret fairly simple graphs because of the “pattern-matching” heuristic. According to this theory people will wrongly extrapolate a trend because it matches a pattern, which leads them to the wrong conclusions about what actions to take.
When you look at predictions of past, present and predicted future greenhouse emissions, you often see two points: the emissions of greenhouse gases (going up) and the total emissions concentration (also going up). Because of the pattern-matching heuristic, most people think reducing emissions will also reduce total emissions, but that’s not correct because the greenhouse effect is cumulative. Every year, with more greenhouse gas emitted, the effect gets stronger. Like water going into a bathtub, you need to actually remove something from the system in order to lower the level; just reducing the rate of additions doesn’t stop the water/temperature rising.
The challenge here isn’t really a technical one; we don’t need to add lots of bells, whistles and interactions to our graphs. The problem is in the brain, and how it’s applying existing knowledge to create a model of the world from the data available. So what could you do to make sure people gain an accurate understanding of the world from a data visualisation.
Add language and guidance alongside your graph.
One quick intervention you can make is to provide warning and guidance which helps users focus on the most important parts of a visualisation. Harold et al (2015) applied this to a set of graphs related to climate change, and found that adding some guidance about correctly identifying long-term trends in datasets with large short-term volatility helped “more detailed internal representations of the data to be formed”.
Focus on the default state.
Following the idea of the “default heuristic”, you find that many people choose to stay with a default option if it is offered (Gigerenzer and Brighton, 2009) whether that’s related to options available at dinner or options provided by an interactive data visualisation. The New York Times Data Lab, for example, reported that only a maximum of 15% of visitors looking at their news pieces will interact with the accompanying visualisations, which allow people to play with data and find their own story.
(This is a write-up of the talk I gave at INCH Munich on March 11 )medium.com
With that in mind, the first view of a graph is often the most important part of the design. If you focus on telling a clear story from the default, you’ll achieve much more impact than if you spend your energy worrying about all the filters, buttons and other configurations that only 15% of your audience are going to use.
Think about the ink.
Edward Tufte is one of the most famous visualisation designers and his idea of the data-ink ratio still influences the way many people approach visualisation design. In the Visual Display of Quantitative Information he wrote about the primary importance of communicating data — the “data ink” — and making sure there’s as little distraction from that as possible. Gridlines, shading and other non-essential ‘ink’ should be removed from your visualisation as far as possible. But pay attention that you don’t take this too far, or you can end up making the data unintelligible again. Talking with people and refining the design based on comments from potential users is an essential part of the process, as always.
There’s a neat example I always recall in the gif below that shows the balance between stripping back all the non-essential ink, then re-adding emphasis to deliver a compelling story.
Histograms are great for comprehension.
I couldn’t write a post about data visualisation without paying respect to the humble histogram. Researchers asked 162 people working on climate change in the UK and Germany to interpret the same data on a number of different type of graphs and then rate their favourite (Lorenz et al 2015). The histogram was interpreted most accurately and was often picked as the first choice for using yourself, or for showing others.
A lot of this can be explained when looking at the Basic Tasks model (Carswell, 1992) a framework to understand the varying effectiveness of different ways of interpreting data. The datapoint’s position on a scale or the datapoint’s length (bar size) are very effective means of interpreting a dataset — and are both qualities that histograms excel at.
But sometimes something more abstract can be more effective.
There’s a number of different ways to show data about our world, or about a specific process. If you want to talk about the nutrient cycle, for example, would it be better to animate the whole process and cram in lots of detail or create a series of static images of the key sections. A paper by Mary Hegarty (2011) includes a number of examples illustrating that the simpler, more abstract visualisations lead to models being perceived more accurately. One example in that paper is work by Smallman and St John (2005) which indicates that 3D Globes can lead to a number of misunderstandings around distance.
Tversky et al (2002) come to the same conclusion around animated charts, arguing that many animations are either too fast or too complex, reducing the brain’s ability to interpret and memorise the data they’re seeing. However, animation can be a useful tool for provoking emotional responses that aid the brain’s processing of data, such as aesthetic appeal or humour.
In the end… it’s all about the user.
I’ve introduced a few different tricks in this post. They can be used in combination or in isolation, depending on your objectives. Thinking about the ink and focussing on the default state should be treated as general principles though — I think you should take them with you into everything. Whether you choose a visualisation that’s abstract or animated, with or without guidance, histogram or not, is something you should discuss as you go.
But as well as bearing these tips in mind, an understanding of the audience you are designing for should underpin every choice you make. Knowing who they are, their skills, their preferences, and what they want to find out, are all essential for making good decisions about what your visualisation should look like.
If you liked this post, why not take a look at one of our other posts about helping people understand data through visualisations